Demand Side I: Labor Markets, Productivity, and Automation

Session 2 · Mon, Sep 28

POLISCI 158 · AI & Politics · Stanford · Fall 2026

Roadmap

  1. In this session
  2. Questions we’ll tackle
  3. Before class
  4. After class

In this session

We start from a simple political economy recipe: workers exposed to a new technology lose, everyone else enjoys lower prices, and their politics follows. With AI, “exposure” can mean displacement (jobs automated away) or reinstatement (jobs made more productive), and our best measures disagree most about the most exposed jobs. The newest payroll evidence finds no economy-wide job loss so far, but falling employment for young workers in AI-exposed occupations, mostly through less hiring. We then ask when job losses turn into political action, using the China trade shock as a benchmark, and what compensation and organized labor can do for the workers who lose.

Questions we’ll tackle

  • If your exposure is your politics, what does it mean to be “exposed” to AI: substitution or augmentation?
  • Why do AI exposure indexes disagree most about the most exposed jobs, and can we trust layoff announcements instead?
  • Early-career workers in AI-exposed jobs are being hired less, not fired more. Who notices a job that was never offered, and can that loss be mobilized politically?
  • Will AI’s productivity gains show up in workers’ pay?
  • When do we expect a stronger political response: smaller, concentrated job shocks or larger, economy-wide shocks?
  • Does mass sacrifice bring mass compensation, and what role can organized labor play?

Before class

Read in this order

Listen or watch

  • The New Jobs AI Will Create, The AI Daily Brief, Nathaniel Whittemore (May 10, 2026), 31 min. The case for reinstatement: how cheaper AI-enabled services could create demand for new kinds of human work.

Explore

  • AI Economic Indicators, Stanford Digital Economy Lab (ongoing, updated monthly). The Canaries Dashboard extends the paper above with monthly payroll data; the other dashboards track AI adoption and macroeconomic signals. Pick one and note what it can and cannot tell you.
  • California AI-Unemployment Tracker, California Policy Lab and California Employment Development Department (2026). Unemployment insurance claims by AI exposure, education, and region; compare the statewide picture with the Bay Area.
  • JobsLost.ai (2026). A tracker of AI-attributed layoffs compiled from company announcements and news reports; ask what counting announcements misses.
  • The Productivity–Pay Gap, Economic Policy Institute (data 1948–2026). Productivity and typical workers’ pay grew together until the late 1970s and then diverged; ask whether AI would change that pattern.

After class

Key takeaways

  • The baseline model is material self-interest: exposed workers lose jobs or wages and oppose the change, while those who are not exposed enjoy lower prices. The same recipe has been used for trade and globalization.
  • AI exposure cuts two ways. Full automation of a job substitutes for workers (fewer jobs, lower wages); partial automation can augment them (more jobs, higher productivity and wages). Which one dominates is an open empirical question.
  • Measurement is hard. Exposure indexes disagree more about highly exposed jobs, layoff announcements are noisy, and California’s tracker finds no statewide surge in AI-related claims but rising claims among college-educated workers in high-exposure jobs and in the Bay Area.
  • The Stanford Digital Economy Lab’s payroll data show no economy-wide displacement through June 2026, but employment of 22- to 25-year-olds in AI-exposed occupations stands 19% below where it would be had it kept pace with their less-exposed peers. The gap comes mainly from less hiring, not more firing, and it is concentrated where AI is used to automate rather than complement work. The authors present these as descriptive facts, not causal estimates.
  • Preferences alone do not pass policy; job losses need mobilization to produce a response. After the China shock, each $1,000 of import exposure per worker cut annual household wage and salary income per working-age adult by about $549, while government transfers rose by only about $58 per person, just $3.65 of it through unemployment insurance and Trade Adjustment Assistance.
  • Historically, large compensation followed mass sacrifice. Organized labor is one channel for AI: the Writers Guild won AI protections in its 2023 contract, and public approval of unions is high, but union membership is at a historic low.

Go further

About this deck

Adapted from POLISCI 157 Session 2A (Jun 29, 2026).

Curated by Simone Paci, Stanford University. Drafted with AI assistance (Claude) from the instructor’s course materials and reviewed by the instructor before publication; every external link is checked before each update.

Demand Side II: Firms, Platforms, and Lobbying →